Identification and Analysis of Influence in Social Networks: User Centric Approach

Abstract

Social network is now becoming an indispensable part of the society. A large number of social networks play a vital role in the dissemination of information regarding various products, services, socioeconomic events etc. In addition, they influence the products one buys, places one visits; many a times whom one votes, events one attends etc. The countless ways in which social network affect lives, makes it important to understand its structure and investigate it further to make it an effective tool for various useful applications. This research focuses on the influence maximization problem, which aims to fetch information propagation initiators, for the vast spread of information. However, picking the correct propagation initiators may not suffice for an effective optimal solution. Other aspects such as network structure and influence among users, have to be investigated. In this work, a new model to map user’s role during information propagation is presented. Along with this model, a holistic approach for influence maximization taking into consideration three aspects of social networks; i) network structure, ii) influence probability and iii) top influential users, is designed. The first task is to fetch the sub set of users, who actively take part in the spread and adoption of information and opinions. This aspect is closely associated to target selection problem. The exponential and rapid growth of social networks in terms of users is a major challenge for its analysis. Due to the huge run time of popular influence maximization solutions, like the Greedy algorithm, distance, degree etc., it is difficult to evaluate its effectiveness in the enormous social networks. This research work addresses the scalability issue by reducing the social networks to smaller key components. This pruned networkcomprises of probable adopters and spreaders of information, thus, making information propagation effective. User influence plays an important role in social network analysis including influence maximization. Therefore, second task is to estimate user influence in social networks. In practice, influence probabilities have significant implications for applications such as viral marketing, poll prediction, political campaigns, recommendation system etc. Yet, predicting influence probabilities has not received significant research attention. In this research, Influx approach is devised to estimate user influence. This is further used to design a new variant of the independent cascade model, namely Influx-IC model. This model is used to predict the spread of information that is initiated by influential users. The final stage is to fetch the top influential users in the social network, who can influence a vast population to adopt the information. To achieve this task, a new centrality metric is proposed. Based on this metric, two new heuristics are designed. Further, the heuristics employed with the estimated value of influence is used to predict the information diffusion in the social networks. In the previous works the solution to influence maximization has been explored on either models, heuristics or estimating parameters such as influence. This research sets itself apart from its predecessors by identifying vital aspects that play an important role in estimating information diffusion. Further, this research proposes a holistic approach that solves influence maximization by amalgamating aspects of social network pruning, user influence and fetching top influential users. The combination of these aspects provide an effective and viable solution to predict the information diffusion in the social networks in the real world.